A Hybrid Approach for Path Planning and Execution for Autonomous Mobile Robots

Path planning and execution are very important tasks for autonomous mobile robots. In the environment considered in this study, first the path must be planned from a source point to a destination point. Next, the control navigation is executed from a set of possible actions. Several works based on metric map can be found in the literature. This paper address an approach in which a path as a sequence of actions is executed by a robot from some decision points in the environment, so that the robot does not need to constantly update its localization in termos of (x, y) coordinates. The decision points are not known in advance and the robot must identify them during its navigation. A hybrid approach is proposed so that a genetic algorithm can find the sequence of reactive behaviors the robot should execute to reach the destination and a pattern recognizer is used to identify the decision points. The robot needs to automatically recognize the decision points to select a new action for each situation. Experiments were performed in the Player/Stage simulator and the hybrid approach achieved promising results regarding the path planning and execution under the conditions defined.